US6907507B1

Movatterモバイル変換

Info

Publication number: US6907507B1
Application number: US10/326,432
Authority: US
Inventors: Oleg Kiselev; Anand A. Kekre; John A. Colgrove
Original assignee: Veritas Operating Corp
Current assignee: Arctera Us LLC
Priority date: 2002-12-19
Filing date: 2002-12-19
Publication date: 2005-06-14
Also published as: US7089385B1

Abstract

Disclosed is a method and apparatus for tracking in-progress writes to a data volume and a copy thereof using a multi-column bit map. The method can be implemented in a computer system and, in one embodiment, includes creating a data volume in a first memory, and creating a copy of the data volume in a second memory. In response to the computer system receiving a request to write first data to the data volume, the computer system switches the state of first and second bits of a map entry in a memory device, wherein the state of the first and second bits are switched using a single write access to the memory device.

Description

BACKGROUND OF THE INVENTION

Many businesses rely on large-scale data processing systems for storing and processing their data. Insurance companies, banks, brokerage firms, etc., rely heavily on data processing systems. Often the viability of a business depends on the reliability of access to data contained within its data processing system. As such, businesses seek reliable ways to consistently protect their data processing systems and the data contained therein from natural disasters, acts of terrorism, or computer hardware and/or software failures.

Businesses must be prepared to eliminate or minimize data loss and recover quickly with useable data after an unpredicted event such as a hardware or software failure in the data processing system. When an unexpected event occurs that causes loss of data, businesses often recreate the data using backup copies of their data made of magnetic storage tape. Restoring data from a backup tape is typically a time-consuming process that often results in a substantial loss of business opportunity. Business opportunity is lost because the data processing system cannot process new data transactions while data is being recreated from backup copies. Further, restoration from tape usually involves loss of data. For most businesses, this kind of data loss and down time is unacceptable. Mirrored data volume and virtual point-in-time (PIT) backup technologies offer better solutions for full and accurate restoration of business critical data. Mirrored data volume and virtual PIT technologies not only minimize or eliminate data loss, but also enable rapid recovery of a data processing system when compared to conventional bulk data transfer from sequential media.

FIG. 1 illustrates an exemplarydata processing system10 that employs mirrored data volume and virtual PIT backup technology.Data processing system10 includes ahost node12 coupled to data storage systems14-18. Data storage systems14-18 include data memories24-28, respectively. As will be more fully described below,data memory24 stores a primary data volume,data memory26 stores a virtual PIT backup copy of the primary data volume, anddata volume28 stores a mirrored copy of the data volume.Data processing system10 shown in FIG.1 and its description herein should not be considered prior art to the invention disclosed and/or claimed.

Host node

12 may take form in a computer system (e.g., a server computer system) that received and processes requests to read or write data to the primary data volume stored withinmemory24. In response to receiving these requests,host node12 generates read or write-data transactions for reading or writing data to the primary data volume withindata memory24. It is noted thathost node12 is capable of accessing each of the memories24-28 or memory internal to hostnode12 for reading or writing data to the data volumes stored therein.

Each of the data memories24-28 may include several memory devices such as arrays of magnetic or optical discs. The primary data volume maybe distributed across several memory devices ofmemory24. Likewise, the virtual PIT copy of the primary data volume is distributed across several memory devices ofdata memory26, and the mirrored copy of the primary data volume is distributed across several memory devices ofdata memory28. Each of the data memories24-28 may include n_maxphysical memory blocks into which data can be stored. More particularly,data memory24 includes n_maxmemory blocks allocated byhost node12 for storing data of the primary data volume,data memory26 includes n_maxmemory blocks allocated to store the virtual PIT backup copy of the primary data volume, anddata memory28 includes n_maxmemory blocks allocated to store the mirrored copy of the primary data volume. Corresponding memory blocks in data memories24-28 are equal in size. Thus,memory block1 ofdata memory24 is equal in size tomemory block1 of

data memories

26 and28.

Host node

12 creates the virtual PIT backup copy of the primary data volume according to the methods described in copending U.S. patent application Ser. No. 10/143,059, filed May 10, 2002 entitled “Method and Apparatus for Creating a Virtual Data Copy,” and U.S. patent application Ser. No. 10/254,753, filed Sep. 25, 2002 and entitled “Method and Apparatus for Restoring a Corrupted Data Volume,” each of which is incorporated herein by reference in entirety. These references describe howhost node12 reads and writes data to the virtual PIT copies. Additionally, the above referenced applications describe how the virtual PIT backup copy can be converted to a real PIT backup copy using a background copying process executed byhost node12.

Host node

12 creates the virtual PIT backup copy of the primary data volume when host node receives or internally generates a command to create a PIT copy of the primary data volume. Initially (i.e., before any virtual PIT copy is created in data memory26)data memory26 contains no data. Whenhost node12 creates the first virtual PIT backup copy inmemory26,host node12 creates a pair of valid/modified (VM) maps, such as

maps

30 and32 shown in FIG.2.FIG. 2 also shows athird map34 which will be more fully described below.

VM maps

30 and32 corresponds to the primary data volume and the virtual PIT copy thereof, respectively, stored within

data memories

24 and26, respectively. Hence,

VM maps

30 and32 will be referred to asprimary VM map30 and PIT backupcopy VM map32.

Each of the

VM maps

30 and32 include n_maxentries, each entry having two bits. Each entry ofprimary VM map30 corresponds to a respective block ofdata memory24, while each entry of the virtual PIT backupcopy VM map32 corresponds to a respective block ofdata memory26. The first and second bits in each entry of

VM maps

30 and32 are designated V_nand M_n, respectively. V_nof each entry, depending on its state (i.e., logical 0 or logical 1), indicates whether block n of the associated memory contains valid data. V_nof theprimary VN map30 indicates whether a corresponding memory block n inmemory24 stores valid data. For example, when set to logical 1, V₂ofVM map30 indicates thatblock2 ofdata memory24 contains valid data of the primary data volume, and when set to logical 0, V₂of theVM map30 indicates thatblock2 ofdata memory24 contains no valid data of the primary data volume. V_nof the PIT backupcopy VM map32 indicates whether memory block n inmemory26 stores a valid point-in-time copy of data in block n ofmemory24. For example, V₂ofVM map32, when set to logical 1, indicates thatblock2 ofmemory26 contains a valid copy of data that existed inblock2 ofmemory24 at the time the pit backup copy was first created or at the time the PIT backup copy was last refreshed. Copending U.S. application Ser. No. 10/326,427, filed Dec. 19, 2002, entitled “Instant Refresh of A Data Volume” describes one method of refreshing a PIT backup copy and is incorporated herein by reference in entirety. V₂ofVM map32, when set to logical 0, indicates thatblock2 ofmemory26 does not contain a valid copy of data. The V_nbit of PIT backupcopy VM map32 is used to determine when data in block n ofmemory24 is to be copied to block n ofmemory26. More particularly, whenhost node12 generates a write-data transaction for writing data to block n ofmemory24,host node12 checks the status of V_nin PIT backupcopy VM map32. If V_nis set to logical 0, then the PIT backup copy inmemory26 lacks a valid copy of data of block n inmemory24. Before data can be written to block n ofmemory24 in accordance with the write-data transaction, data in block n ofmemory24 must first be copied to block n ofmemory26. If, on the other hand, V_nis set to logical 1, data can be written to block n ofmemory24 in accordance with the write-data transaction without first copying data to block n ofmemory26.

M_nin each entry ofVM map30 and PITcopy VM map32, depending upon its state, indicates whether data within a respective block n of the corresponding memory is modified (or new) since the time the PIT backup copy was first created or last refreshed. For example, when set to logical 1, M₃of theprimary VM map30, indicates thatblock3 ofdata memory24 contains data which has been modified since the time the PIT backup copy was first created or last refreshed. When set to logical 0, M₃of the primary VM map of30 indicates that data has been modified or written toBlock3 ofdata memory24 since the time the PIT backup copy was first created or refreshed.

When

VM maps

30 and32 are first created byhost node12, each entry of PITVM map32 is set to logical 0 thus indicating thatdata memory24 contains no valid or modified data. For purposes of explanation, it is presumed that each block ofdata memory24 contains valid data of the primary volume. Accordingly, VM of each entry withinprimary VM map30 is initially set to logical 1. Lastly, M_nof each entry inprimary VM map30 is initially set to logical 0.

As noted above,data memory28 stores a mirrored copy of the primary data volume. A mirrored copy is considered a real time copy of the primary data volume. Eachtime host node12 writes data to the primary data volume stored inmemory24 via a write-data transaction, host node also generates a transaction to write the same data to the mirrored copy stored withinmemory28. Thus, the mirrored copy, as its name implies, closely tracks the changes to the primary data volume. The mirrored copy withinmemory28 provides an immediate backup data source in the event thatdata storage system14 fails as a result of hardware and/or software failure. Ifdata storage system14 suddenly becomes unusable or inaccessible,host node12 can service client computer system read or write requests using the mirrored volume withindata memory28.

Host node

12 is also subject to hardware and/or software failure. In other words hostnode12 may crash. Whenhost node12 recovers from a crash,host node12 expects contents of the primary data volume to be consistent with the contents of the primary data volume. Unfortunately,host node12 may crash after a write-data transaction is completed todata memory24, but before the mirrored copy can be updated with the same data of the write-data transaction. If this is to happen, the contents of the primary data copy stored inmemory24 will be inconsistent with the mirrored copy inmemory28 whenhost node12 recovers from its crash.

Host node

12 uses the dirty region (DR) map34 shown inFIG. 2 to safeguard against inconsistencies between

memories

24 and28 after crash recovery.DR map34 includes n_maxentries corresponding to the n_maxmemory blocks in

data memories

24 and28. Each entry includes one bit which, when set to logical 0 or logical 1, indicates whether respective memory blocks in

data memories

24 and28 are the subject of an in-progress write-data transaction. For example,host node12 sets DR₂to logical 1 beforehost node12 writes data tomemory block2 in

data memories

24 and28. Setting or clearing a bit inDR map34 requires a write (i.e., an I/O operation) to memory that storesDR map34. After the write transaction is completed tomemory block2 of

data memories

24 and28,host node12 switches the state of DR₂back to 0 using another write operation.

Ifhost node12 crashes, the contents ofDR map34 is preserved. After recovering from the crash,host node12 copies the contents of memory blocks indata memory24 to respective memory blocks indata memory28, or vice versa, for each respective DR_nbit set to logical 1 inDR map34. After these memory blocks are copied,host node12 is ensured that the primary data volume ofdata memory24 is consistent with the mirrored copy ofdata memory28.

Host node

12 is burdened by having to generate two separate write operations to updateVM map30 andDR map34 with each write-data transaction to the primary data volume. More particularly, with each write-datatransaction host node12 must generate a first write operation to updateVM map30 and a second write operation to updateDR map34. Additionally, a substantial amount of time is needed to complete separate write operations toVM map30 andDR map34. To illustrate,FIG. 3 is a flow chart illustrating operational aspects of modifying or writing data of the primary volume withindata memory24 in accordance with a write-data transaction. More particularly, whenhost node12 receives a request to write data to the primary volume from one of the computer systems,host node12 generates a write-data transaction to write data to block n ofmemory24. Instep42,host node12

accesses VM map

32 to determine whether V_nis set to logical 1. If V_nof theVM map32 is set to logical 0, thenhost computer12, as shown instep44, copies the contents of block n ofmemory24 to the corresponding block n inmemory26. Thereafter, instep46

host computer

12 sets V_nof theVM map32 to logical 1. Beforehost computer system12 writes or modifies data of block n ofmemory24 that stores a portion of the primary volume and block n ofmemory28 that stores a portion of the mirrored copy,host computer system12 must set M_nand DR_nofprimary VM map30 andDR map34, respectively, to logical 1. It is noted that M_nand DR_nare in separate tables, and thus have separate addresses within memory. Accordingly,host computer system12 must generate separate writes or IO functions to set M_nand DR_nto logical 1. The time it takes, for example, for the read/write head (not shown) to travel across a magnetic disc to reach the appropriate sector that stores M_nofVM map30, followed by the time it takes for the magnetic read/write to travel across the magnetic disc to reach another sector that stores DR_nof theDR map34, may be substantial. Once M_nand DR_nare set to logical 1 in

steps

50 and52,host node12 writes data to block n inmemory24 as shown instep54 and writes data to block n ofmemory28.

SUMMARY OF THE INVENTION

BRIEF DESCRIPTION OF THE DRAWINGS

The present invention may be better understood, and its numerous objects, features, and advantages made apparent to those skilled in the art by referencing the accompanying drawings.

FIG. 1 is a block diagram of a data processing system;

FIG. 2 is a block diagram of VM and DR maps created by the host node shown inFIG. 1;

FIG. 3 is a flow chart illustrating operational aspects of writing or modifying data in the primary data volume and the mirrored copy thereof, after creation of a virtual PIT backup copy of the primary data volume;

FIG. 4 is a block diagram of a data processing system employing one embodiment of the present invention;

FIG. 5 is a block diagram of VM/DR and VM maps created by the host node shown inFIG. 4, and;

FIG. 6 is a flow chart illustrating operational aspects of writing or modifying data in the primary data volume and the mirrored copy thereof of the data processing system shown inFIG. 4, after creation of a virtual PIT backup copy of the primary data volume.

The use of the same reference symbols in different drawings indicates similar or identical items.

DETAILED DESCRIPTION

The present invention relates to an apparatus and method for tracking in-progress write I/O's through the use of multi-column bit maps.FIG. 4 illustrates (in block diagram form) relative components of adata processing system60 employing one embodiment of the present invention.Data processing system60 is similar todata processing system10 of FIG.1.Data processing system60 includes ahost node62 coupled to data storage systems64-68. It is noted that the term “coupled devices” should not be limited to devices coupled directly together by a communication link. Two devices may be coupled together even though communication occurs through an intervening device.

Data storage systems64-68 include memories74-78, respectively.Memory74 stores one or more primary data volumes. For purposes of explanation,memory74 stores one primary data volume, it being understood that the present invention will not be limited thereto.Memory76 stores a point-in-time (PIT) backup copy of the primary data volume whether real or virtual, whilememory78 stores a mirrored copy of the primary data volume.

The primary data volume (and the PIT backup and mirrored copies thereof) is a collection of files that store data. While it is said that files store data, in reality data is stored in physical blocks ofmemory74 which are allocated to the files byhost node62.Host node62 may take form in a computer system (e.g., a server computer system) that includes a datastorage management system70 in the form of software instructions executing on one or more data processors (not shown). Datastorage management system70 may include a file system (not shown) and a system (not shown) for managing the distribution of volume data across several memory devices. Volume Manager™ provided by Veritas Corporation of Mountain View, Calif., is an exemplary system for managing the distribution of volume data across several memory devices.Host node62 generates read or write-data transactions in response tohost node62 receiving requests from client computer systems to read or write data to the primary data volume.

As noted, memories74-78 store the primary data volume, the PIT backup copy of the primary data volume, and the mirrored copy of the primary data volume, respectively. Data memories74-78 may take form in one or more dynamic or static random access memories, one or more arrays of magnetic or optical data storage disks, or combinations thereof. Data memories74-78 should not be limited to the foregoing hardware components; rather, data memories74-78 may take form in any hardware, software, or combination of hardware and software in which data may be accessed and persistently stored. Data memories74-78 may take form in a complex construction of several hardware components operating under the direction of software. The memories may take form in mirrored hardware. It is further noted that the present invention may find use with many types of redundancy/reliability systems. For example, the present invention may be used with Redundant Array of Independent Disks (RAID) systems. Moreover, the present invention should not be limited to use in connection with the host node of a data storage network. The present invention may find use in a storage switch or in any of many distinct appliances that can be used with a data storage system.

Each of the data memories74-78 includes n_maxphysical blocks of memory into which data can be stored. It is noted that any or all of the memories74-78 may have more than n_maxmemory blocks. However, the first n_maxblocks in memories74-78 are allocated byhost node62 for storing the primary data volume, the PIT backup copy of the primary data volume, and the mirrored copy of the primary data volume, respectively. Corresponding memory blocks in memories74-78 are equal in size. Thus, memory blocks1 ofmemory74 is equal in size tomemory blocks1 in

memories

74 and76. Each of the memory blocks withinmemory74 may be equal in size to each other. Alternatively, the memory blocks inmemory74 may vary in size.

Host node

62 can access each of the blocks in memories74-78 in response to generating read or write-data transactions. The read or write-data transactions are generated byhost node62 in response tohost node62 receiving read or write requests from client computer systems coupled thereto. The primary data volume stored inmemory74 is the “working” volume ofdata processing system60. Thus, when client computer systems request a read or write access to data, the access is directed to the primary data volume. The PIT backup copy of the primary data volume acts as a safeguard against data corruption of the primary data volume due to a software or operator error. The mirrored copy withinmemory78 provides a alternative source to data in the event thatdata storage system64 is rendered inaccessible tohost node62 due to hardware and/or software failure.

Host node creates a virtual PIT backup copy of the primary data volume by creating primary VM/DR map80 and aPIT VM map82 shown within FIG.5. These maps may be stored within one of the memories74-78, in a memory withinhost node62 or in a memory (not shown) attached to hostnode62. The contents of each of the

maps

80 and82 is accessible byhost node62 using a read or write operation.

Maps

80 and82, like

maps

30 and32 described above, include n_maxentries. Each entry of primary VM/DR map80 corresponds to a respective blocks of

data memories

74 and78, while each entry of theVM map82 corresponds to a respective block ofdata memory76.VM map82 is substantially similar to map32, while VM/DR map80 is substantially similar to a combination ofVM map30 andDR map34.

Each entry in primary VM/DR map80 consists of three bits designated V_n, M_n, and DR_n. V_nin the primary VM/DR map, depending on its state, indicates whether a corresponding block inmemory74 contains valid primary volume data. For example, when set to logical 1, V₂of the primary VM/DR map80 indicates thatblock2 ofmemory74 contains valid primary volume data, and when set to logical 0, V₂of primary VM/DR map80 indicates thatblock2 ofdata memory74 contains no valid primary volume data. V₂ofVM map82, when set to logical 1, indicates thatblock2 ofmemory76 contains a valid copy of data that existed inblock2 ofmemory74 at the time the PIT backup copy was first created or at the time the PIT backup copy was last refreshed. V₂ofVM map82, when set to logical 0, indicates thatblock2 ofmemory76 does not contain a valid copy of data.

M_nin each entry of primary VM/DR map80, depending upon its state, indicates whether data has been written to or modified in the corresponding block n ofmemory74 since the last time the PIT backup copy of the primary data volume was created or refreshed. For example, when set to logical 1, M₃of primary VM/DR map80 indicates that data has been written to block3 ofmemory74 since the last time PIT backup copy was refreshed. Likewise, when set to logical 0, M₃of the primary VM/DR map80 indicates that data has not been written or modified inblock3 of thememory74 since the last time a PIT backup copy was refreshed.

DR_nin each entry, depending on its state, indicates whether the corresponding block n ofmemories74 and79 are the subject of an in-progress write-data transaction. For example, when set to logical 1, DR₅of VM/DR map80 indicates thatmemory block5 of

memories

74 and78 are the subject of an in-progress write-data transaction. When set to logical 0, DR₅of primary VM/DR map80 indicates thatmemory block5 of

memories

74 and78 are not the subject of an in progress write transaction.

Multiple bits in entries of primary VM/DR map80 are can be set byhost node62 using a single write operation to memory that stores VM/DR map80. In one embodiment, multiple bits of an entry can be updated with a single write operation since the bits are stored adjacent to each other in the memory that stores VM/DR map80. More particularly, host node can write a multi-bit binary value to an entry of VM/DR map80 using a single write operation thereby updating or changing one or more bits of the entry. For example, suppose V₃, M₃, and DR₃ofentry3 of VM/DR map80 are initially set to logical 1, logical 0, and logical 0, respectively. In other words,entry3 of VM/DR map80 stores binary value “100” where the most significant bit (i.e., the left most bit) corresponds to M₃and the least significant bit (i.e., the right most bit) corresponds to DR₃. Suppose further thathost node62 generates a write-data transaction for writing data to block3 ofmemory74. Before or afterhost node62 writes data to block3,host node62 can write the binary value “111” toentry3 of VM/DR map80 thereby changing the state of M₃and DR₃from logical 0 to logical 1 (V₃is maintained at logical 1 even though V₃may be overwritten) using a single write operation. Incontrast host node12, described in the background section above, requires separate operations to update M₃ofVM map30 and DR₃ofDR map34 where one operation changes the state of M₃ofVM map30 while the other operation changes the state of DR₃ofDR map34. It is noted thathost node62 in this example could write the binary value “x11” toentry3 of VM/DR map80 instead of the binary value “111.” In this alternative embodiment, V₃could be masked from the write operation so that it is not overwritten by the binary value “x.” Indeed, any one of the V_n, M_n, or DR_nbits can be masked when a multi-binary value is written to VM/DR map.

In yet another alternative embodiment, multiple multi-bit entries in VM/DR map80 can b modified byhost node62 via a single write operation. For example, supposehost node62 generates a write-data transaction for writing data to block4 ofmemory74 after host node writes data to block3 ofmemory78. Further suppose that

entries

3 and4 of VM/DR map80 store binary values “111” and “100,” respectively, at thetime host node62 generates the write-data transaction for writing data to block4 ormemory74. DR₃is set to logical 1 thus indicating thatblock3 in

memories

74 and76 are the subject of a write data transactions when, in fact, they are not. VM/DR map80 can be updated, before or after data is modified inblock4 ofmemory78 in accordance with the new write-data transaction, by writing binary values “110” and “111” to

entries

3 and4, respectively, via a single write operation since

entries

3 and4 rather than two write operations where the first operation writes binary value “110” toentry3 and the second operation writes binary value “111” toentry4.

Entries

3 and4 can be updated with a single write operation since

entries

3 and4 ofmap80 are adjacent to each other in the memory that stores VM/DB map80. It is noted that V₃, V₄, and M₃are not updated in the sense that the value of these bits do not change after the single write operation to VM/DR map80. After the single write operation, DR₃is properly updated to logical 0 thus indicating thatblock3 in

memories

74 and78 are no longer subject to a write-data transaction, and M₄is updated to logical 1 thus indicating that data ofblock4 inmemory74 has been modified.

Operational aspects of tracking in-progress writes is best understood with reference to the flow chart ofFIG. 6 which shows operational aspects of one embodiment of the present invention. Tracking in-progress writes begins afterhost node62 receives a request to write data to the primary data volume.Host node62 identifies the block n ofmemory74 that is the target of the received write request.Host node62 generates a write-data transaction for writing data to block n instep90.Host node62, instep92, accessesVM map82 to determine the state of V_n. If V_nis set to logical 0, then hostnode62 copies the data of block n inmemory74 to block n ofmemory76 and subsequently sets V_nofVM map82 to logical 1 with a write operation to the memory that storesmap82. If V_nofVM map82 is determined to be set to logical 1 instep92 or after V_nis set to logical 1 instep96, the process proceeds to step100 were host node sets M_nand DR_nbits of VM/DR map80 using a single write operation to memory that stores VM/DR map80. Thereafter in

steps

102 and104, block n of

memories

64 and68 are updated with the write data of the request instep90. It is noted that

steps

102 and104 can be reversed in order or performed in parallel. Eventually, when block n of

memories

64 and68 have been updated with the new data,host node62, in synchronous fashion, accesses entry n of VM/DR map80 and sets the DR_nbit to logical 0, thus indicating that the data write transaction to block n of

memories

74 and78 have completed and block n ofmemory74 stores valid and modified data.

Because only one write operation is needed to update the M_nand DR_nbits of primary VM/DR map80, the amount of processing overhead onhost node62 is reduced when compared tohost node12 in FIG.1. Moreover, because write operations ofstep100 shown inFIG. 6 requires a single access to a physical address within memory that stores the primary VM/DR map80, the amount of time needed to perform M_nand DR_nupdating is substantially reduced.

Claims

1. A method comprising:

creating a data volume in a first memory;

creating a copy of the data volume, wherein the copy is created in a second memory;

a computer system receiving a request to write first data to the data volume;

switching the state of a pair of bits in a memory device in response to the computer system receiving the request to write first data to the data volume, wherein the state of the pair of bits are switched using a single write access to the memory device.

2. The method ofclaim 1 further comprising allocating memory of the memory device for a map, wherein the map comprises an entry, wherein the entry comprises the pair of bits.

3. The method ofclaim 2 further comprising:

writing the first data to a memory block in the first memory;

writing the first data to a memory block in the second memory;

switching the state of one or the pair back to its original state after writing the first data to the memory blocks of the first and second memory.

4. The method ofclaim 3:

wherein a first bit of the pair, when switched to a first or second state, indicates that data has been written to the first memory after allocation of the memory for the map;

wherein a second bit of the pair, when switched to a first or second state, indicates that the first and second memories will be subject to a write operation.

5. The method ofclaim 1 further comprising:

creating a map in the memory device, wherein the map comprises a plurality of entries, each entry comprising first, second and third bits;

wherein each entry corresponds to a respective memory block in the first memory;

wherein each first bit, when set to a first or second state, indicates whether its respective memory block of the first memory stores valid data;

wherein each second bit, when set to a first or second state, indicates whether its respective memory block of the first memory stores data which has been modified since creation of the data volume copy in the second memory;

wherein each third bit, when set to a first or second state, indicates whether its respective memory block of the first memory will be subject to a write operation, and;

wherein the pair of bits are the second and third bits, respectively, of one entry of the map.

6. A computer readable medium comprising instructions executable by a computer system, wherein the computer system performs a method in response to executing the instructions, the method comprising:

creating a data volume in a first memory coupled to the computer system;

creating a copy of the data volume, wherein the copy is created in a second memory coupled to the computer system;

switching the state of a pair of bits in a memory device in response to the computer system receiving a request to modify first data of the data volume, wherein the state of the pair of bits are switched using a single write access to the memory device.

7. The computer readable medium ofclaim 6, wherein the method further comprises allocating memory of a memory device for a map, wherein the map comprises an entry, wherein the entry comprises the pair of bits.

8. The computer readable medium ofclaim 7 wherein the method further comprises:

modifying the first data, wherein the first data is stored in a memory block in the first memory;

modifying data in the second memory;

switching the state of one bit of the pair back to its original state after modifying data in the second memory.

9. The computer readable medium ofclaim 7:

wherein one bit of the pair, when switched to a first or second state, indicates that data in the first memory was modified after allocation of the memory for the map;

wherein the other bit of the pair, when switched to a first or second state, indicates that the first and second memories are subject to write operation.

10. The computer readable medium ofclaim 7 wherein the method further comprises:

creating a map in the memory device, wherein the map comprises a plurality of entries, each entry comprising first, second, and third bits;

wherein each first bit, when set to a first or second state, indicates whether its respective memory blocks of the first memory stores valid data;

11. A computer readable medium comprising instructions executable by a computer system, wherein the computer system performs a method in response to executing the instructions, the method comprising:

creating a data volume in first memory coupled to the computer system;

creating a mirror of the data volume, wherein the mirror is created in a second memory coupled to the computer system;

creating a map in a memory device, wherein the map comprises a plurality of entries, wherein each entry comprises first and second bits;

wherein each first bit, when set to a first or second state, indicates whether data has been written to a respective block of the first memory since allocation of memory for the map;

wherein each second bit, when set to a first or second state, indicates whether respective blocks in the first and second memories will be subject to separate in-progress write operations.

12. The computer readable medium ofclaim 11 wherein each of the plurality of entries comprises a third bit, wherein the third bit, when set to a first or second state, indicates whether respective blocks in the first memory contains valid data.

13. A method implemented in a computer system, the method comprising:

creating a data volume, a portion of which is stored in a first memory block;

creating a mirror of the data volume, a portion of which is stored in a second memory block;

creating a map in a memory device, wherein the map comprises a plurality of entries, wherein one of the plurality of entries comprises first and second bits;

the computer system receiving a request to write first data to the first memory block;

switching the state of the first and second bits after the computer system receives the request;

writing the first data to the first and second memory blocks;

switching the state of the second bit after the first data is written to the second memory block.

14. A computer readable medium comprising instructions executable by a computer system, wherein the computer system performs a method in response to executing the instructions, the method comprising:

creating a data volume, a portion of which is created in a first memory block;

creating a copy of the data volume, a portion of which is created in a second memory block;

creating a map in a memory device, wherein the map comprises a plurality of entries, wherein one of the plurality of entries comprises first and second bits, wherein the first bit, depending on its state, indicates whether data has been written to the first memory block since allocation of memory for the map, wherein the second bit, depending on its state, indicates that the data contents of the first memory block should be overwritten with the data contents of the second memory block or vice versa;

switching the states of the first and second bits with a single write access to the memory device.

15. A data processing system comprising:

a first memory for storing a data volume;

a second memory for storing a mirrored copy of the data volume;

a host node coupled to receive a request to write new data to or modify existing data of the data volume;

wherein the host node is configured to switch the state of first and second bits in a memory device in response to the host node receiving the request to write new data to or modify existing data of the data volume, wherein the state of the first and second bits are switched using a single write access to the memory device.

16. A data processing system comprising:

a first memory for storing a data volume;

a second memory for storing a mirrored copy of the data volume;

a means coupled to receive a request to write new data to or modify existing data of the data volume;

wherein the means is configured to switch the state of first and second bits in a memory device in response to the means receiving the request to write new data to or modify existing data of the data volume, wherein the state of the first and second bits are switched using a single write access to the memory device.